Siirry pääsisältöön

Information seeking

Artificial intelligence and Information Retrieval

This page discusses artificial intelligence in information retrieval for higher education studies. Artificial intelligence and its applications are developing rapidly, and this guide will be continuously updated. Please regularly check this page for new information and follow the latest developments in AI to stay informed. 

Key concepts

  • Artificial intelligence refers to a machine's ability to use skills traditionally associated with human intelligence, such as reasoning, learning, planning, or creating.
  • Machine learning is a branch of artificial intelligence rooted in statistics. Machine learning methods are developed based on the given data without needing program-specific rules. They improve their performance in a specific task by accumulating more data.
  • A large language model (LLM) is based on the probabilities of the occurrence of words and strings of words. It predicts the continuation of the text input or produces text as requested.
  • Generative AI combines the power of machine learning, deep learning, and AI. Responding to a request or a prompt, it can create original content such as text, videos, audio, code, or images. The generative AI model is trained with big data and user feedback, and based on this, it can create new innovative outputs.

Xamk's AI policies

Artificial intelligence instructions for students and teachers

Artificial intelligence instructions for teachers

Artificial intelligence and licensed materials

  • Xamk's licensed materials, e.g. scientific articles or e-books, may not be fed into any language models or other AI tools. CC licences do not override this rule.

ChatGPT

ChatGPT is a service from OpenAI that offers two different versions of language models: ChatGPT (free) and GPT4 (paid). The app can be found at https://chat.openai.com/chat. You need an OpenAI account to use it.


You interact with ChatGPT by entering text, such as a question or statement (a prompt), and it responds to the user's input. In other words, you are conversing with the generative AI.


ChatGPT is trained using massive amounts of text and data. It generates responses based on the texts it has analysed based on a statistical model. In other words, it’s comparable to predictive text on a mobile phone. The free version of ChatGPT knows nothing by itself, and it can't be used as a source of information. Nor can it assess the reliability of the text it produces, and it can create false facts. OpenAI has not disclosed any information about the data used to train ChatGPT. It's also worth remembering that AI models are notoriously biased.


N.B! Open AI updated its paid ChatGPT-4 in September 2023, and it can now take advantage of the internet.
 

Orientation to the topic

  • You can use ChatGPT for inspiration when brainstorming and formulating a research question. Be prepared for the fact that it may include prejudiced assumptions in the discussion. Remember that the free version was trained with data until 2021 and has no knowledge of the latest research results. Therefore, your information might be out of date. 

Search preparation

  • ChatGPT can help you find good keywords for your information search. It's a good idea to ask ChatGPT to complete its list of keywords several times rather than settle for the first answer. However, it does not replace the need to familiarise yourself with industry glossaries and terminology. Always check the latest terms used by experts in your field.
  • ChatGPT can help you identify tools or databases suitable for your topic. Please note that it may not provide a comprehensive answer regarding field-specific databases. Comprehensive information on information sources is available from the library's Databases A-Z list.
  • ChatGPT is not able to formulate search phrases for searching for scientific information. If you ask it to create search phrases, you will probably have to edit and reformat them.

How to perform a search

  • ChatGPT cannot help you with the actual information retrieval. 
    • It does not know how to search for information systematically and transparently.
  • ChatGPT can tell you how the database works. However, we recommend that you check the instructions directly from the database of your choice. The functionality may have changed since 2021, and ChatGPT's response may be outdated.

Evaluation of results

  • You cannot use ChatGPT to evaluate search results. Evaluating the quality of information is an important human skill, based on both objective and subjective criteria. As a researcher or student, you need to decide whether the source is relevant to your work.
  • However, you can use ChatGPT to clarify and translate difficult texts, which is good for condensing or reformatting text. However, remember source criticism, i.e., not believing everything and that the service may present biased views and the translation may not be accurate. 

Citation and references

  • ChatGPT is unable to provide correct sources for the texts it writes. It can produce a bibliography when requested, but usually, the sources are incorrect or completely fabricated.
  • Also, ChatGPT cannot generate bibliographies according to citation patterns. Reference managers like Zotero support many citation styles and are good tools for managing your references.

When using AI applications, always consider how they process the data you enter. As a rule, all material entered into AI services is stored outside the EU/EEA and cannot be removed. For example, do not enter personal data, business secrets, and confidential information. It is a breach of confidentiality. 
 

Modified from the chart ”Is it safe to use ChatGPT for your task?” By Aleksander Tiulkanov CCBY

Microsoft Copilot

Microsoft Copilot (formerly Bing Chat Enterprise) is a conversational AI application (GPT-4 language model and DALL-E 3 image model). Copilot can also search for information directly from the Internet and create images based on written descriptions or prompts.
Copilot is available for Xamk's staff and students.

The data processed in Copilot is not stored by artificial intelligence and is not used to train artificial intelligence. Conversations with Copilot are automatically deleted at the end of the chat session. For more information, see Copilot's privacy policy.

Artificial intelligence in the search for scientific information

This section presents applications in which artificial intelligence is mainly used to search for international scientific information. Xamk does not have organisational licenses for these services. Remember data security and exercise caution when using these services (see below ). 
Please also note that services may become chargeable or disappear at any time.
 

Keenious

  • Free and paid versions. The paid version, Keenious Plus, has more features than the free version. When logged in, you get acce.ss to more features.
  • A tool that utilises existing text as a basis for information retrieval. It searches for articles on the same topic based on a ready-made article or other text entered into it. In addition to article search, Keenious produces a list of topics that it interprets as related to the text entered into it. Based on this, the user can refine their search.
  • In addition to the website, the tool can also be used with a Word or Google Docs plug-in

Semantic Scholar

  • It is free to use, but signing up gives you access to more features.
  • The search operates in the same way as Google Scholar.
  • Displays article citations and tries to find not only the most cited articles but also the most relevant articles. It also makes article recommendations based on articles selected by the user.
  • Suitable for those interested in filtering and organising search results and saving articles for new recommendations.

Elcit

  • Getting started with the service is free but requires a login. After a certain amount of use, you will need to pay to continue to use Elicit. The price is determined by the amount of use.
  • The tool searches for and summarises scientific articles based on a given research question or article. Unlike traditional search services, Elicit does not indicate the number of search results on the search results page but displays the seven most relevant articles.
  • Elicit also offers other academic work tools; it helps formulate a research question, analyse articles, and create abstracts.

Litmaps

  • Free and paid versions are available, but there are differences in the number of articles and maps.
  • It can create a bibliography map with data from one or several publications using BibTeX/RIS files.
  • Litmaps displays relationships between scientific articles in a dynamic network graph. It makes suggestions for articles based on links between them, which can help you find articles outside of a traditional keyword search.

Research Rabbit

  • Usage requires an account.
  • The tool can help you find new connections between different phenomena and researchers.
  • You start to use the tool by adding one good article to the collection. The tool generates recommendations based on items in your collection using headlines, abstracts, and other metadata.
  • It helps you visualise the networks formed between publications. You can move forward on the map and create new networks of publications with authors and sources.

Ethical use of artificial intelligence in information retrieval

  • Generative AI is a tool that can help us in everyday life, work, or study. As with any tool, using them ethically, critically and appropriately is important.
  • Can AI be used in the task? If so, where? And how should it be reported?
  • Misinformation is either incomplete or false information that is given inadvertently. It is not intended to mislead deliberately.
  • Disinformation is deliberately shared misinformation that can be motivated by, among other things, political and social influence, economic gain, and damage.
  • AI tools may not disclose their sources, nor may they use sources that meet the requirements for scientific writing. In some cases, sources have proven to be fictitious or inaccurate. If you're using a tool to produce some elements of your work, the person reviewing the work needs to know which bits are yours and which aren’t.
  • AI-based search tools might give the impression that they can do scientific research easily and quickly. However, this may mean the student fails to learn essential scientific research skills such as information retrieval, critical thinking, source evaluation, and problem-solving.
  • For example, to use ChatGPT, you need to create an account. The terms of use for your account permit the service to gather your input and personal data, which is a privacy issue. AI tools may require you to provide a phone number or other personal information.
  • You should always be careful about what information you share when creating an account online.
  • Be sure to carefully review the program's privacy policy and terms of use. Remember that when using these tools, your conversations are typically recorded and used by the service in the future.
  • In some cases, AI tools can access online material without the authors' permission. Compliance with intellectual property rights is a core part of good scientific practices and the ethical conduct of research, so it is essential to take a critical view of using such tools in your work.

Design AI-driven prompts

AI works with prompts. AI is not a mind reader, so you should carefully describe your request and elaborate it if necessary. Do not be afraid to challenge the answer you are given or ask for additional information. This can help you produce the desired results. Designing and constructing well-written prompts is a skill and a prerequisite for successfully using artificial intelligence tools.

The Digital Toolkit of the Tampere higher education community describes the basics of creating prompts.

AI prompts | Digital toolkit | Tampere Universities (tuni.fi)

Student Checklist

  • Discuss with your course teacher or seminar instructor well in advance whether AI applications (and which ones) are allowed for information retrieval in your work.
  • Research different AI applications before using them, including what they can and cannot do to support your information search and retrieval, and compare the results.
  • Consider what you can and cannot use an AI application for in information retrieval.
  • Get to know the AI application you use: who runs it and why? Is it commercial or open source? How does this affect the results? Are there any biases to be aware of? Do you need to report this or not?
  • Thoroughly and critically checking your sources is still your job. Be critical of the results: make comparative samples from different sources that provide information and check, e.g., references provided by artificial intelligence and the references they receive from databases. Check whether the references it has produced actually exist. Sometimes, they can also be artificially generated.
  • See instructions for using an AI application as a support tool for text production: https://libguides.xamk.fi/c.php?g=692278&p=4961691#s-lib-ctab-15582581-19 

 

This Artificial intelligence in information retrieval guide is based on Tampere University Library's guideInformation Searching Guide: Information searching and AI: https://libguides.tuni.fi/information-searching-guide/information-searching-ai

Saavutettavuusseloste | Accessibility statement